kevin/linux
1
Fork 0
Code

um: fix time-travel syscall scheduling hack

Browse Source 
The schedule() call there really never did anything at
least since the introduction of the EEVDF scheduler,
but now I found a case where we permanently hang in a
loop of -ERESTARTNOINTR (due to locking.) Work around
it by making any syscalls with error return take time
(and then schedule after) so we cannot hang in such a
loop forever.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
This commit is contained in:
Johannes Berg 2024-08-27 16:05:01 +02:00 committed by Richard Weinberger
parent ae0dc67c25
commit 381d2f95c8
1 changed files with 20 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
arch/um/kernel/skas/syscall.c
View File

@ -12,23 +12,13 @@
#include <sysdep/syscalls.h>
#include <linux/time-internal.h>
#include <asm/unistd.h>
#include <asm/delay.h>
void handle_syscall(struct uml_pt_regs *r)
{
struct pt_regs *regs = container_of(r, struct pt_regs, regs);
int syscall;
/*
* If we have infinite CPU resources, then make every syscall also a
* preemption point, since we don't have any other preemption in this
* case, and kernel threads would basically never run until userspace
* went to sleep, even if said userspace interacts with the kernel in
* various ways.
*/
if (time_travel_mode == TT_MODE_INFCPU ||
time_travel_mode == TT_MODE_EXTERNAL)
schedule();
/* Initialize the syscall number and default return value. */
UPT_SYSCALL_NR(r) = PT_SYSCALL_NR(r->gp);
PT_REGS_SET_SYSCALL_RETURN(regs, -ENOSYS);
@ -41,9 +31,25 @@ void handle_syscall(struct uml_pt_regs *r)
goto out;
syscall = UPT_SYSCALL_NR(r);
if (syscall >= 0 && syscall < __NR_syscalls)
PT_REGS_SET_SYSCALL_RETURN(regs,
EXECUTE_SYSCALL(syscall, regs));
if (syscall >= 0 && syscall < __NR_syscalls) {
unsigned long ret = EXECUTE_SYSCALL(syscall, regs);
PT_REGS_SET_SYSCALL_RETURN(regs, ret);
/*
* An error value here can be some form of -ERESTARTSYS
* and then we'd just loop. Make any error syscalls take
* some time, so that it won't just loop if something is
* not ready, and hopefully other things will make some
* progress.
*/
if (IS_ERR_VALUE(ret) &&
(time_travel_mode == TT_MODE_INFCPU ||
time_travel_mode == TT_MODE_EXTERNAL)) {
um_udelay(1);
schedule();
}
}
out:
syscall_trace_leave(regs);